A QSAR model of benzoxazole derivatives as potential inhibitors for inosine 5`-monophosphate dehydrogenase from Cryptosporidium parvum

Cryptosporidium parvum is the common enteric protozoan pathogen causing cryptosporidiosis in human. Available drugs to treat cryptosporidiosis are ineffective and there is yet no vaccine against C. parvum. Therefore, it is of interest to design an improved yet effective drug against C. parvum. Here, we docked benzoxazole derivatives (collected from literature) with inosine 5`- monophosphate dehydrogenase (IMPDH) from Cryptosporidium parvum using the program AutoDock 4.2. The docked protein - inhibitor complex structure was optimized using molecular dynamics simulation for 5 ps with the CHARMM-22 force field using NAMD (NAnoscale Molecular Dynamics program) incorporated in visual molecular dynamics (VMD 1.9.2) and then evaluating the stability of complex structure by calculating RMSD values. NAMD is a parallel, object-oriented molecular dynamics code designed for high-performance simulation of large biomolecular systems. A quantitative structure activity relationship (QSAR) model was built using energy-based descriptors as independent variable and pIC50 value as dependent variable of fifteen known benzoxazole derivatives with C. parvum IMPDH protein, yielding correlation coefficient r2 of 0.7948. The predictive performance of QSAR model was assessed using different cross-validation procedures. Our results suggest that a ligand-receptor binding interaction for inosine 5`-monophosphate dehydrogenase using a QSAR model is promising approach to design more potent inosine 5`-monophosphate dehydrogenase inhibitors prior to their synthesis.

are needed. The sequencing of the genomes of Cryptosporidium parvum revealed a highly streamlined anabolic metabolism with potential choke points that might be exploited in drug design [5]. One such vulnerability lies in the pathway that supplies purine nucleotides for the synthesis of DNA and RNA. Like all protozoan parasites, Cryptosporidium is incapable of de novo purine synthesis and relies on salvage of purines from the host [5]. Adenosine is converted into guanine nucleotides in a streamlined pathway that relies on inosine 5`-monophosphate dehydrogenase (IMPDH) catalyzing the conversion of inosine 5`-monophosphate (IMP) to xanthosine 5`-monophosphate (XMP) [6].

©2016
In this study, we docked experimentally verified 15 benzoxazole-based inhibitors having inhibitory value IC50 in nM with C. parvum IMPDH using AutoDock 4.2, which resulted in energy-based descriptors. Molecular dynamics (MD) simulation studies of inhibitor -protein complex were performed and after that, we have built quantitative structure activity relationship (QSAR) model using Multiple Linear Regression.

Methodology: Protein target structure
The 3D coordinates of the crystal structure of the catalytic domain of the Inosine 5`-monophosphate dehydrogenase from cryptosporidium parvum (PDB Id: 4IXH) was retrieved from Protein Databank (http://www.rcsb.org/) and is shown in Figure 1. This is used as a target model for flexible docking. The structure was optimized using the chimera tool [7].

Inhibitors dataset
Fifteen benzoxazole derivatives with known pIC50 were obtained from Gorla et al. (2013) [8]. The 3D structures of known 15 inhibitors were downloaded in .sdf format from pubchem compound database. They were later converted in .pdb format with the help of open babel [9] tool. All the ligands were subjected to energy minimization using the HyperChem software [10].

Molecular docking
Docking of fifteen benzoxazole derivatives screened from literature against C. parvum IMPDH structure were done using molecular docking program AutoDock [11]. Gasteiger charges are added to the ligand and maximum 6 numbers of active torsions are given to the lead compounds using AutoDock tool [12]. Kollman charges and the solvation term were added to the protein structure. The Lamarckian genetic algorithm implemented in Autodock was used for docking.

Molecular dynamics simulations
Molecular dynamics simulations were done using the NAMD (NAnoscale Molecular Dynamics program; v2.7) graphical interface module [13] incorporated visual molecular dynamics (VMD 1.9.2) [14]. The protein-ligand complex was immersed in the center of a 50 Å box of water molecules where all water molecule atoms (H-O-H) were closer than 1.5 Å and a CHARMM (Chemistry at HARvard Macromolecular Mechanics) 22 parameter file for proteins and lipids; phi and psi cross-term map correction were used in the force field for complexes. For the minimization and equilibration of complex in the water box, we assumed force-field parameters excluding scaling of 1.0 Å and a cutoff of Coulomb forces with a switching function starting at 12 Å, reaching zero at a distance of 10 Å, ending at 14 Å with a margin of 3.0 Å, and all atoms, including those of hydrogen, were illustrated explicitly. A protein structure file (psf) stores structural information of the protein, such as various types of bonding interactions. The psf was created from the initial pdb and topology files using psfgen package of VMD. After running psfgen,two new files were generated protein pdb and protein psf and by accessing PSF and PDB files; NAMD generated the trajectory DCD file. After the simulations, the results were analyzed in VMD by calculating the Root mean square deviation (RMSD) of the complex using rmsd tcl source file from the Tk console and finally rmsd.dat was saved and accessed in Microsoft office excel 2007.
Several cross-validation procedures were adopted to assess the predictive performance of the QSAR model. In leave-one-out strategy (LOOCV), one molecule was removed from the dataset as a test compound and the remaining 14 molecules were used to build the model. This process was repeated 15 times with each inhibitor as a test molecule.      Table 3. Further, the docked complexes were analyzed through Python Molecular Viewer software [15] for their interaction studies and were shown in Figure 3. Thus from the Complex scoring and binding ability it's deciphered that these compounds are promising inhibitors for IMPDH protein. MD simulation is a well-known theoretical technique and is mainly used for evaluating the stability of any predicted 3D model. Therefore, the constructed 3D model of protein-ligand complexes were processed for MD simulation for a 5 ps timescale with Langevin dynamics to control the kinetic energy, temperature, and/or pressure of the system. The RMSD values of complexes contain alpha carbon atoms, and all atoms were calculated by taking structure with reference conformation points. The RMSD values of complex versus time were shown in Figure 4. Relationship between experimental and predicted pIC50 values of Benzoxazole derivatives is shown in Figure 2.

Conclusion:
A QSAR model using pIC50 values for fifteen known benzoxazole derivatives binding with C. parvum IMPDH protein as dependent variable and molecular docking based predicted pIC50 with a correlation coefficient r2 is 0.7948 was reported.